I can't understand the result in README.md file, "It turn out that Tucker decomposition yields lower accuracy loss than CP decomposition in my experiments, so the results below are all from Tucker decomposition."

[wangwang110]

"It turn out that Tucker decomposition yields lower accuracy loss than CP decomposition in my experiments, so the results below are all from Tucker decomposition."

what means "Tucker decomposition yields lower accuracy loss than CP decomposition" ?

and can the project be used in other cnn models?

[Yuncheng Jiang]

"It turn out that Tucker decomposition yields lower accuracy loss than CP decomposition in my experiments, so the results below are all from Tucker decomposition."

what means "Tucker decomposition yields lower accuracy loss than CP decomposition" ?

and can the project be used in other cnn models?

it means after fine-tuning the tucker decomposd model, the prediction accuracy will recover to higher score than the cp decompsed model. in a word, tucker decomposition is better than cp decomposition.

Both tucker decomposition and cp decomposition is to achieve higher inference speed by lossing a certain model accuracy. But the tucker can find a better balance between the speed and accuracy by the VBMF ranking method